FAME: Towards Factual Multi-Task Model Editing

This is the official code repository for the EMNLP 2024 paper FAME: Towards Factual Multi-Task Model Editing.

Setup Instructions

To run the code, follow these steps:

git clone https://github.com/BITHLP/FAME.git
cd fame
conda create -n fame python==3.11.5
conda activate fame
pip install -r requirements.txt

Running the Model

cd ./code/model
python default_batch_qa.py

Evaluating the Results

Once the model has finished running, follow these steps to calculate accuracy:

cd ../scripts
python cal_acc.py

Demo Information

This is just a demo to showcase our FAME dataset and the basic workflow of SKEME. However, there are some network-related bugs in the online query part of SKEME. We are working on fixing these bugs and will upload the complete version of the code soon.

TODO:

• Upload the full code
• Upload the process for RQs